研究目的
To restore 3D single-photon Lidar images obtained under challenging scenarios such as imaging multilayered targets or through obscurants by developing a new algorithm that accounts for Poisson statistics, non-local spatial correlations, and collaborative sparsity.
研究成果
The proposed NR3D algorithm effectively restores 3D single-photon Lidar images in challenging conditions, demonstrating robustness to sparse data and high background levels. It shows promise for future applications in multi-modal data fusion and imaging through obscurants like water or fog.
研究不足
The approach assumes the presence of multi-peaks, which may not be highlighted in all scenarios due to space limitations. It relies on manually selected regularization parameters and does not extensively use multi-modal data for weight setting in this implementation.
1：Experimental Design and Method Selection:
The methodology involves designing a convex cost-function that combines Poisson data statistics with regularization terms for non-local spatial correlations and collaborative sparsity. The minimization is solved using the alternating direction method of multipliers (ADMM).
2：Sample Selection and Data Sources:
Synthetic data generated from the Middlebury dataset and real data from a life-sized polystyrene head acquired using a TCSPC system.
3：List of Experimental Equipment and Materials:
A time-correlated single-photon counting (TCSPC) module, laser pulses, and detectors for Lidar imaging. Specific models or brands are not mentioned.
4：Experimental Procedures and Operational Workflow:
Data is acquired by emitting laser pulses and recording photon arrival times to construct histograms. The algorithm processes these histograms to restore depth and reflectivity images.
5：Data Analysis Methods:
Performance is evaluated using signal-to-reconstruction error (SRE) metrics for depth and reflectivity, comparing the proposed NR3D algorithm with classical cross-correlation and RDI-TV methods.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
